A Comprehensive Characterization of Hyper-Morph, Hypo-Morph, and Neo-Morph Mutations in Cancer
Video
Team Information
Team Members
Somnath Tagore, Postdoctoral Researcher Scientist, Department of Systems Biology, CUIMC
Samuel Tsang, Research Assistant Professor, Department of Cell Development and Cancer Biology, Oregon Health and Science University
Gordon Mills, Professor, Department of Cell Development and Cancer Biology, Oregon Health and Science University
Faculty Advisor: Andrea Califano, Clyde '56 and Helen Wu Professor of Chemical Biology (in Systems Biology) and Professor of Biomedical Informatics and Biochemistry and Molecular Biophysics and Professor of Medicine in the Institute for Cancer Genetics, Vagelos College of Physicians and Surgeons
Abstract
Although annotation of genomic mutations is a highly relevant and complex segment of the analysis of sequence-based genomic analyses, currently more than ten million variants lack functional annotation. While computational predictions of variant function are usually integrated into gene-based analyses of rare-variants, there is limited information for assessing variant function in the context of a particular disease. The goal of this study is to assess the mechanistic characterization of context-dependent hyper-, hypo-, and neomorph mutations in cancers. We developed a computational approach to characterize these mutations and assess the functional effects of mutations occurring in the same or different protein domains, and differentiate mutations in terms of their hyper-morph (gain-of-function), hypo-morph (loss-of-function), neutral, or neo-morph (establishing a novel function) potential. To elucidate the functional consequence of such mutations, we based our computational pipeline on inferring the activity of regulatory proteins such as transcription factors (TFs) and co-factors (co-TFs) using the VIPER algorithm. Our analytical pipeline integrates structural and functional information encompassing six topics : 1) structural domains affected by the mutation, 2) the overlap between mutation-specific TF/co-TFs, 3) differential activity signatures and signatures induced by established hyper-morph, hypo-morph and neutral or neo-morph mutations, 4) in vitro data generated by reporter assays, 5) the VIPER-inferred activity of each protein relative to a validated control, and 6) the fraction of proteins in a sample that are not affected by established hyper-morphs and hypo-morphs (candidate neo-morphs).The current repertoire considers 25 TCGA cohorts for 3830 proteins (oncoproteins/tumor suppressors). As an example, we validated our method on the TCGA Breast Cancer dataset (TCGA-BRCA) by predicting with very high confidence several neo-morphic phenotypes, including the previously-described PIK3CAE545, PIK3CAE542, PIK3CAH1047, PIK3CAQ546K and PIK3CAG1049R. Interestingly, PIK3CAE545K, classified previously as a gain-of-function mutation (in one TCGA-BRCA sample) or a loss-of-function mutation (in two other TCGA-BRCA samples), is predicted to be a neo-morph based on our approach. Further validation of these mutations using PIK3CA reporter assays led to the identification of several significant hypo-morphic signals in TP53 mutant samples. We defined this phenomenon as mutational mimicry (i.e., mutations in proteins mimicking those in established oncogenes) and we propose it as a tool for predicting tumor sensitivity/resistance to drugs.
Team Lead Contact
Somnath Tagore: st3179@cumc.columbia.edu
